[molpro-user] Problem with awcvxz-pp basis set?

[Taube, Andrew]

Hi,
I am having a problem with the relatively new aug-cc-pwCVXZ-PP basis sets for iodine and xenon (at least, haven't gone through calculations for bromine and krypton yet). 

When using the basis sets w/o the reweighted core functions on iodine, i.e. 
***,sp~dfmp2~avqz
memory,1000,M
thrdscf=1e-12,thrdscf_max=1e-9
gthresh,throvl=1e-9,throrth=1e-7,energy=1e-8,optgrad=1e-4,optstep=1e-4,optenerg=1e-6
gdirect
basis={
default=avqz,i=avqz-pp
set,jk,context=jkfit
default=avqz,i=AQZVP
}
geomtyp=xyz
symmetry,nosym
orient,noorient
geometry={
7
name
O1 1.7435045780 -0.4531154611 -0.6327326063
H2 1.4572371768 -0.6541823454 0.2702699373
H3 1.6518406105 0.5133564902 -0.6380289538
O4 0.9569201064 -0.2003794367 2.2852703128
H5 0.7392163721 0.6848891838 1.9213759420
H6 1.8455439772 -0.0480272240 2.6233630795
I7 0.9904291790 2.9568307932 0.7420582886
}
set,charge=-1
{df-hf,accu=15,df_basis=jk}

The first iteration of the Hartree-Fock is reasonable and convergence proceeds accordingly:
1PROGRAM * RHF-SCF (CLOSED SHELL)       Authors: W. Meyer, H.-J. Werner


 NUMBER OF ELECTRONS:      23+   23-    SPACE SYMMETRY=1    SPIN SYMMETRY: Singlet 
 CONVERGENCE THRESHOLDS:    3.16E-08 (Density)    1.00E-09 (Energy)
 INTEGRAL THRESHOLDS:       1.00E-11 (Initial)    1.00E-11 (Final)
 MAX. NUMBER OF ITERATIONS:        4
 INTERPOLATION TYPE:            DIIS
 INTERPOLATION STEPS:              2 (START)      1 (STEP)
 LEVEL SHIFTS:                  0.00 (CLOSED)  0.00 (OPEN) 



 
 
 Orbital guess generated from atomic densities. Full valence occupancy:   27
 Coulomb and exchange fitting
 Thresholds for fitting:       THRAO_COUL= 1.0D-14    THRAO_EXCH= 1.0D-12    THRMO= 1.0D-11    THRPROD= 1.0D-11    THRASM= 1.0D-11
                               THR2HLF=    1.0D-11

 Molecular orbital dump at record        2100.2

 ITERATION   DDIFF            GRAD            ENERGY        2-EL.EN.            DIPOLE MOMENTS         DIIS   CPU(IT)  CPU(TOT)  ORB.
    1      0.000D+00      0.000D+00      -446.81647145    802.471989   -1.63668   -3.39929   -0.53083    0      4.6     25.0     start
    2      0.000D+00      0.319D-02      -446.88047029    790.566814   -1.87511   -3.73346   -0.90902    1      4.6     29.6     diag
    3      0.355D-02      0.196D-02      -446.90202272    794.279080   -1.80137   -3.67486   -0.82169    2      4.6     34.2     orth

On the other hand, if I use the reweighted core basis on iodine
***,sp~dfmp2~avqz
memory,1000,M
thrdscf=1e-12,thrdscf_max=1e-9
gthresh,throvl=1e-9,throrth=1e-7,energy=1e-8,optgrad=1e-4,optstep=1e-4,optenerg=1e-6
gdirect
basis={
default=avqz,i=awcvqz-pp
set,jk,context=jkfit
default=avqz,i=AQZVP
}
geomtyp=xyz
symmetry,nosym
orient,noorient
geometry={
7
name
O1 1.7435045780 -0.4531154611 -0.6327326063
H2 1.4572371768 -0.6541823454 0.2702699373
H3 1.6518406105 0.5133564902 -0.6380289538
O4 0.9569201064 -0.2003794367 2.2852703128
H5 0.7392163721 0.6848891838 1.9213759420
H6 1.8455439772 -0.0480272240 2.6233630795
I7 0.9904291790 2.9568307932 0.7420582886
}
set,charge=-1
{df-hf,accu=15,df_basis=jk}

I get nonsense:
1PROGRAM * RHF-SCF (CLOSED SHELL)       Authors: W. Meyer, H.-J. Werner


 NUMBER OF ELECTRONS:      23+   23-    SPACE SYMMETRY=1    SPIN SYMMETRY: Singlet 
 CONVERGENCE THRESHOLDS:    3.16E-08 (Density)    1.00E-09 (Energy)
 INTEGRAL THRESHOLDS:       1.00E-12 (Initial)    1.00E-12 (Final)
 MAX. NUMBER OF ITERATIONS:      100
 INTERPOLATION TYPE:            DIIS
 INTERPOLATION STEPS:              2 (START)      1 (STEP)
 LEVEL SHIFTS:                  0.00 (CLOSED)  0.00 (OPEN) 



 
 
 Orbital guess generated from atomic densities. Full valence occupancy:   27
 Coulomb and exchange fitting
 Thresholds for fitting:       THRAO_COUL= 1.0D-15    THRAO_EXCH= 1.0D-13    THRMO= 1.0D-12    THRPROD= 1.0D-12    THRASM= 1.0D-12
                               THR2HLF=    1.0D-12

 Molecular orbital dump at record        2100.2

 ITERATION   DDIFF            GRAD            ENERGY        2-EL.EN.            DIPOLE MOMENTS         DIIS   CPU(IT)  CPU(TOT)  ORB.
    1      0.000D+00      0.000D+00      -353.34721489    453.526994  -10.02201   46.40416   -9.48511    0      6.0     52.1     start
    2      0.000D+00      0.730D-01      -159.26305208   1795.049533    9.58128 -101.59433    2.81927    1      6.1     58.1     diag
    3      0.118D+01      0.242D+00      -199.81465289   1692.049476    9.56600 -101.42245    2.80492    2      6.1     64.2     diag

I have checked that this is not a density fitting problem. It seems like either the initial guess is very bad or there is something wrong with the underlying basis set. Does anyone have any ideas?

Thanks,
Andrew
---
Andrew G. Taube
D. E. Shaw Research
email: Andrew.TaubeXX_AT_XXDEShawResearch.com